Merge pull request #66 from i-dot-ai/feature/theme_model_changes

Feature/theme model changes

consultation_analyser/consultations/migrations/0003_remove_theme_unique_up_to_question_theme_is_outlier_and_more.py

@@ -0,0 +1,30 @@
+# Generated by Django 5.0.4 on 2024-04-05 14:40
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ("consultations", "0002_alter_theme_options_theme_question_and_more"),
+    ]
+
+    operations = [
+        migrations.RemoveConstraint(
+            model_name="theme",
+            name="unique_up_to_question",
+        ),
+        migrations.AddField(
+            model_name="theme",
+            name="is_outlier",
+            field=models.BooleanField(default=False, editable=False),
+        ),
+        migrations.AlterField(
+            model_name="theme",
+            name="keywords",
+            field=models.JSONField(default=list, editable=False),
+        ),
+        migrations.AddConstraint(
+            model_name="theme",
+            constraint=models.UniqueConstraint(fields=("label", "question"), name="unique_up_to_question"),
+        ),
+    ]

consultation_analyser/consultations/ml_pipeline.py

@@ -53,16 +53,15 @@ def get_answers_and_topics(topic_model: BERTopic, answers_list: List[Dict[str, U
     # Assign topics to answers
     answers_df = topic_model.get_document_info(free_text_responses)
     answers_df["id"] = answers_id_list
-    answers_df = answers_df[["id", "Topic", "Name", "Representation"]]
+    answers_df = answers_df[["id", "Name"]]
     return answers_df
 
 
 def save_themes_to_answers(answers_topics_df: pd.DataFrame) -> None:
     for row in answers_topics_df.itertuples():
         answer = models.Answer.objects.get(id=row.id)
         theme_label = row.Name
-        theme_keywords = row.Representation
-        answer.save_theme_to_answer(theme_label=theme_label, theme_keywords=theme_keywords)
+        answer.save_theme_to_answer(theme_label=theme_label)
 
 
 def save_themes_for_question(question: models.Question) -> None:

consultation_analyser/consultations/models.py

@@ -86,17 +86,28 @@ class Meta(UUIDPrimaryKeyModel.Meta, TimeStampedModel.Meta):
 
 
 class Theme(UUIDPrimaryKeyModel, TimeStampedModel):
+    # Label summarises a topic from a topic model for a given question
     label = models.CharField(max_length=256, blank=True)
     summary = models.TextField(blank=True)
-    keywords = models.JSONField(default=list)
     # Duplicates info in Answer model, but needed for uniqueness constraint.
     question = models.ForeignKey(Question, on_delete=models.CASCADE, null=True)
 
+    # Calculated fields
+    keywords = models.JSONField(default=list, editable=False)
+    is_outlier = models.BooleanField(default=False, editable=False)
+
     class Meta:
         constraints = [
-            models.UniqueConstraint(fields=["summary", "label", "keywords", "question"], name="unique_up_to_question"),
+            models.UniqueConstraint(fields=["label", "question"], name="unique_up_to_question"),
         ]
 
+    def save(self, *args, **kwargs):
+        label_constituents = self.label.split("_")
+        self.keywords = label_constituents[1:]
+        topic_number = label_constituents[0]
+        self.is_outlier = topic_number == "-1"
+        super().save(*args, **kwargs)
+
 
 class Answer(UUIDPrimaryKeyModel, TimeStampedModel):
     multiple_choice_responses = models.JSONField(null=True)  # Multiple choice can have more than one response
@@ -108,12 +119,11 @@ class Answer(UUIDPrimaryKeyModel, TimeStampedModel):
     class Meta(UUIDPrimaryKeyModel.Meta, TimeStampedModel.Meta):
         pass
 
-    def save_theme_to_answer(self, theme_label, theme_keywords):
+    def save_theme_to_answer(self, theme_label):
         question = self.question
         theme, _ = Theme.objects.get_or_create(
             question=question,
             label=theme_label,
-            keywords=theme_keywords,
         )
         self.theme = theme
         self.save()

tests/factories.py

@@ -11,6 +11,15 @@
 default_multiple_choice_options = ["Yes", "No", "Not sure"]
 
 
+def generate_dummy_topic_label():
+    dummy_sentence = faker.sentence()
+    words = dummy_sentence.lower().strip(".")
+    words_joined = words.replace(" ", "_")
+    topic_number = random.randint(-1, 3)
+    output = f"{topic_number}_{words_joined}"
+    return output
+
+
 class FakeConsultationData:
     def __init__(self):
         with open("./tests/examples/questions.yml", "r") as f:
@@ -100,9 +109,8 @@ class Meta:
         model = models.Theme
 
     # TODO - may need to be changed once ML pipeline is in
-    label = faker.sentence()
+    label = factory.LazyAttribute(lambda _o: generate_dummy_topic_label())
     summary = f"Summary: {label}"
-    keywords = label.lower().strip(".").split(" ")
 
 
 class AnswerFactory(factory.django.DjangoModelFactory):

tests/unit/test_ml_pipeline_functions.py

@@ -28,13 +28,10 @@ def test_save_themes_to_answers():
     answers_df = pd.DataFrame(
         {
             "id": [answer1.id, answer2.id, answer3.id],
-            "Topic": [-1, 0, 0],
             "Name": ["-1_x_y", "0_m_n", "0_m_n"],
-            "Representation": [["x", "y"], ["m", "n"], ["m", "n"]],
         }
     )
     ml_pipeline.save_themes_to_answers(answers_df)
     themes_qs = models.Theme.objects.filter(question=question)
     assert themes_qs.count() == 2
     assert "-1_x_y" in themes_qs.values_list("label", flat=True)
-    assert themes_qs.get(label="0_m_n").keywords == ["m", "n"]

tests/unit/test_models.py

@@ -4,21 +4,23 @@
 from tests import factories
 
 
+@pytest.mark.parametrize(
+    "label,expected_keywords,is_outlier", [("0_key_lock", ["key", "lock"], False), ("-1_dog_cat", ["dog", "cat"], True)]
+)
 @pytest.mark.django_db
-def test_save_theme_to_answer():
+def test_save_theme_to_answer(label, expected_keywords, is_outlier):
     question = factories.QuestionFactory(has_free_text=True)
     answer = factories.AnswerFactory(question=question, theme=None)
-    keywords = ["key", "lock"]
-    label = "0_key_lock"
     # Check theme created and saved to answer
-    answer.save_theme_to_answer(theme_label=label, theme_keywords=keywords)
-    theme = models.Theme.objects.get(keywords=keywords, label=label)
-    assert theme.keywords == keywords
+    answer.save_theme_to_answer(theme_label=label)
+    theme = models.Theme.objects.get(keywords=expected_keywords, label=label)
+    assert theme.keywords == expected_keywords
     assert theme.label == label
+    assert theme.is_outlier == is_outlier
     assert answer.theme.label == label
     # Check no duplicate created
-    answer.save_theme_to_answer(theme_label=label, theme_keywords=keywords)
-    themes_qs = models.Theme.objects.filter(keywords=keywords, label=label)
+    answer.save_theme_to_answer(theme_label=label)
+    themes_qs = models.Theme.objects.filter(keywords=expected_keywords, label=label)
     assert themes_qs.count() == 1